Skip to content

Dataset · leduckhai

Vietnamese medical dataset for automatic speech recognition (ASR)

leduckhai/VietMed

Vietnamese audio recordings and transcriptions for training and evaluating medical ASR systems.

What's inside

The dataset contains 16 hours of annotated Vietnamese medical speech, along with 1,000 hours of unlabeled medical audio and 1,200 hours of general-domain recordings. The annotation on Hugging Face features 1 to 10 thousand audio-text examples. The set was presented at LREC-COLING 2024 as the largest open corpus for medical ASR.

What you can do with it

  • Training and fine-tuning (LoRA) ASR models for Vietnamese speech recognition
  • Pre-training acoustic models on unlabeled medical audio data
  • Testing transcription systems in narrow medical terminology
Tasks
Automatic speech recognition
Languages
Vietnamese
Size
1K–10K rows
Downloads this month
1.4K
Likes
28

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("leduckhai/VietMed")